High-resolution tree canopy mapping for New York City using LIDAR and object-based image analysis

TL;DR: Converting high-resolution LIDAR and imagery into usable information is a nontrivial exercise, requiring significant processing time and labor, but an expert system–based combination of OBIA and manual review was an effective method for fine-scale canopy mapping in a complex urban environment.

Abstract: Urban tree canopy is widely believed to have myriad environmental, social, and human-health benefits, but a lack of precise canopy estimates has hindered quantification of these benefits in many municipalities. This problem was addressed for New York City using object-based image analysis (OBIA) to develop a comprehensive land-cover map, including tree canopy to the scale of individual trees. Mapping was performed using a rule-based expert system that relied primarily on high-resolution LIDAR, specifically its capacity for evaluating the height and texture of aboveground features. Multispectral imagery was also used, but shadowing and varying temporal conditions limited its utility. Contextual analysis was a key part of classification, distinguishing trees according to their physical and spectral properties as well as their relationships to adjacent, nonvegetated features. The automated product was extensively reviewed and edited via manual interpretation, and overall per-pixel accuracy of the final map was 96%. Although manual editing had only a marginal effect on accuracy despite requiring a majority of project effort, it maximized aesthetic quality and ensured the capture of small, isolated trees. Converting high-resolution LIDAR and imagery into usable information is a nontrivial exercise, requiring significant processing time and labor, but an expert system–based combination of OBIA and manual review was an effective method for fine-scale canopy mapping in a complex urban environment. © 2012 Society of Photo-Optical Instrumentation Engineers (SPIE). [DOI: 10.1117/1.JRS.6.063567]